FIRE RISK ASSESSMENT IN GERMANY - PROCEDURE, DATA, RESULTS -

Transcription

1 International Conference Nuclear Energy in Central Europe 2000 Golf Hotel, Bled, Slovenia, September 11-14, 2000 FIRE RISK ASSESSMENT IN GERMANY - PROCEDURE, DATA, RESULTS - H.P. Berg Bundesamt für Strahlenschutz (BfS) Willy-Brandt-Str. 5, D Salzgitter, Germany BWillecke@bfs.de 1 ABSTRACT The recommended approach for a quantitative fire risk assessment to be applied within periodic safety reviews of nuclear power plants in Germany starts with a screening process providing critical fire zones and is followed by a quantitative analysis using a standard event tree with elements for fire initiation, ventilation of the room, fire detection, fire suppression, and fire propagation. In a final step, the fire induced frequency of initiating events, the main contributors and the calculated hazard state frequency for the fire event are determined. For that purpose, a comprehensive data base is needed which has been developed in particular for active fire protection measures. As an example results of one fire PSA are reported. 1 INTRODUCTION In the past, most of the engineering work in designing fire protection measures in German nuclear power plants (NPPs) has been performed on a deterministic basis. Moreover, the use of deterministic fire risk analysis is current practice in Germany to review the fire protection state of operating NPP. It should be underlined that these reviews have led to comprehensive backfitting and upgrading measures including structural fire protection measures (e.g. fire barriers) as well as the active fire detection, alarm and extinguishing features and administrative fire protection measures (for manual fire fighting) resulting in significant improvements in fire safety, in particular in case of a NPP built to earlier standards. However, as it can be seen from other areas the probabilistic approach provides different insights into design and availability of systems and components and supplements the results from deterministic analyses. Thus, probabilistic considerations have been taken into account for decision making on a case-by-case basis. A more comprehensive fire risk assessment is recommended in the frame of periodic safety reviews in Germany.

2 H.P.Berg page 2 of REGULATORY GUIDANCE FOR FIRE RISK ASSESSMENT The PSA Guide contains reference listings of initiating events for NPP with PWR and BWR respectively which have to be checked plant specifically with respect to applicability and completeness. Plant internal fires are included in these listings. Detailed instructions are provided in the technical documents on PSA methods [1] and PSA data [2] which are shortly reported. These technical documents have been developed by a working group of technical experts chaired by the BfS (Bundesamt für Strahlenschutz - Federal Office for Radiation Protection). The German regulatory guidance as it is available now has been restricted in scope to comprise only applications where sufficient practical experience is available and a reasonable consensus between the parties involved is achieved. However, the PSA working group is still in force to further develop and amend the technical documents, also with regard to fire specific aspects Screening analysis The screening process to identify critical fire zones is an important first step within a fire risk assessment. Such a screening analysis should not be so conservative that an unmanageable number of fire scenarios remains for the detailed quantitative analysis. However, it must be ensured that all relevant areas are investigated within the quantitative analysis. The systematic check of all rooms/room pairs of the plant can be done in two different ways: The critical fire zones can be identified in a qualitative (qualitative screening) or in a quantitative process (screening by frequency). The qualitative screening allows - due to the introduction of appropriate selection criteria - the determination of critical fire zones with a limited effort. Applying the quantitative screening method, the critical fire zones are identified in a simplified event tree analysis. The determination of critical fire zones starts with the identification of all rooms for which at least one of the following criteria is fulfilled: (L) fire load higher than 25 kwh/m 2, (S) room contains safety related equipment or cables of such equipment, (O) room contains operational equipment, or sensing/control equipment of the reactor protection, or power limit control system, or cables of such equipment with the potential that a fire caused damage may lead to a plant transient/initiating event or to a manually operated scram (Operational equipment is not considered), (V) In case that a fire causes an unintentional opening of a safety valve or of the main steam bypass leading to a loss of coolant accident or a main steam line leak, this fire zone will be classified as a critical fire zone. In a first step, those rooms are identified for which the first three criteria (L), (S) and (O) are simultaneously fulfilled. These rooms will be identified as essential fire zones /rooms. In a next step, for those rooms for which two out of these three criteria are fulfilled, adjacent rooms are checked to identify pairs of rooms that fulfil all three criteria. The critical fire zones/rooms and pairs of rooms are selected based on the further criterion that the fire leads to a safety related initiating event.

3 H.P.Berg page 3 of 8 In the case that the quantitative screening process is applied, the hazard state frequency will be calculated in a simple but conservative analysis for each room with PSA related components and components leading to initiating events after the occurrence of a fire. The event tree analysis will be carried out for fire zones/rooms or room pairs with fire loads > 25 kwh/m 2. Only two elements are taken into account in this event tree analysis: The fire occurrence frequency and the conditional unavailability of the safety related equipment to mitigate the initiating event. All other branches of the event tree, like fire detection, fire suppression and fire spread to adjacent rooms are not considered. All PSA related equipment within the room is assumed to be damaged (probability of damage equal 1.0). Rooms are screened out, if the product of the fire frequency and the conditional unavailability of the safety related equipment is less than 1 % of the total sum of these products. The sum of those neglected contributions shall not exceed 5 % Quantitative analysis For the quantitative part of the fire risk assessment a standard event tree has been developed with nodes for fire initiation, ventilation of the room, fire detection and suppression, both in the pilot fire phase and the fully developed fire phase, as well as fire propagation. This standard event tree must be adapted to each critical fire zone or room. For each critical fire zone/room the following results are obtained: frequency and nature of fire initiating events, list of damaged equipment, categorised corresponding to different damage states, and damage frequencies. If a complete plant specific PSA is available, for each initiator the fire induced frequency will be summed up and specified as input to the corresponding system event tree of the level 1+ PSA. Additionally, the damage state of the equipment has to be introduced into the fault trees. The plant hazard state frequency is calculated for each transient as the sum of the single event core melt frequencies. The total plant hazard state frequency is obtained by summarising the contributions of all transients. The requirement to use only qualified PSA codes has also to be fulfilled for the fire PSA. Moreover, validated fire simulation models and codes have to be used in case of deterministic fire hazard analysis and probabilistic fire risk assessments. 3 3 DATA BASE In order to perform a quantitative fire risk assessment, a basic data base must be established which should, e.g., include initiating event frequencies, reliability data for all fire protection measures, fire barriers, etc. Detailed information is needed on ignition sources, detection and extinguishing systems, manual fire fighting, stationary fire suppression systems; further information on secondary effects, safety consequences, analysis of the cause of the event and corrective measures, etc. would be helpful. It should be underlined that plant specific data are to be used as far as possible. As one contributor to fire specific PSA input data, reliability data for the active fire protection measures are required for the application in the fire specific event tree analysis.

4 H.P.Berg page 4 of 8 These data needed to be estimated are unavailabilities per demand or failure rates per hour of plant operation for those components or systems belonging to the active fire protection means. Active fire protection measures to be studied include all the fire detection equipment, that means all types of fire detectors including their power supplies, the alarm panels and boards, fire doors and dampers and the stationary fire extinguishing systems including the extinguishing media supplies. In Germany, for two different types of reactors (PWR and BWR) unavailabilities and failure rates were estimated. These reliability data are given in tables 1 and 2 and are updated values of those provided in [3,4] by processing additional information. The scattering factor k given in these tables is correlated to the failure rates. The data on potential failures or unavailabilities per demand of the respective fire protection measures were gained from the plant specific documentation of inspection and maintenance. The assessment whether the detected findings are estimated as failures or only as deficiencies or deteriorations requires a deep insight in the plant specific operating conditions for the fire protection means and needs careful engineering judgement. Table 3 gives an overview of the generic failure rates for active fire protection features revealed from the plant specific data estimated. Further detailed information on plant specific and generic estimation of unavailabilities and failure rates for active fire protection measures can be found in [4]. Most of the estimated values for the technical reliability of active fire protection measures represent realistic data for the failure rates of these systems and components in German NPP. They can be applied in the frame of PSA studies instead of conservative data from past nuclear specific reliability studies as well as instead of generic values available from data of the insurance companies for the reliability of comparable fire protection means in nonnuclear industry, both being available up to the time being. The data have been adapted in the PSA data document [2]. 4 4 EXAMPLE OF A RECENT FIRE PSA In the following the fire PSA performed for the Grafenrheinfeld 1300 MW e PWR nuclear power plant meeting the requirements set in the PSA Procedure Guide is shortly described [5]. The fire PSA was performed in two stages (see Figure 1). The first stage covered fires assumed limited to one fire zone and the second stage covered fires which propagated to adjacent fire zones as a result of a failure of the fire protection systems. The obtained core damage frequency due to fire is /a. Compared with the core damage frequency of the initiating events of the Full Power PSA of /a; this is a contribution of 12 %.

5 H.P.Berg page 5 of 8 Table 1: Plant specific reliability data estimated for active fire protection features in a German BWR reference plant Active fire protection feature Fire alarm boards: detection drawers detection lines Fire detectors: automatic press button Inspection perio d 3m, 1a 3m, 1a Scatter ing factor k Failure rate λ (t) [1/h] Unavailability per deman d a 1a Fire dampers 3m, 1a, 3a Fire doors 1a Dry sprinkler extinguishing system (total failure) 6 m, 1a, 5a Dry sprinkler extinguishing system: automatic actuation failure only 6m, 1a, 5a Wet sprinkler extinguishing system 6m, 1a, 5a Gas extinguishing systems (CO 2 ) 6m Stationary fire pumps 1m, 1a Wall hydrants 6m, 1a 7, Table 2: Plant specific reliability data estimated for active fire protection features in a German PWR reference plant Active fire protection feature Fire alarm boards: detection drawers detection lines Fire detectors: automatic press button Inspection perio d 3m 3m Scattering factor k Failure rate λ (t) [1/h] Unavailability per deman d a 1a Fire dampers 6m Fire doors 1a Dry sprinkler extinguishing system (total failure) 3m, 1a Dry sprinkler extinguishing system: automatic actuation failure only 3m, 1a Stationary fire pumps 1m, 1a Wall hydrants 1a

6 H.P.Berg page 6 of 8 Table 3: Generic reliability data estimated for active fire protection features in German NPP Active fire protection feature Failure rate λ (t) [1/h] Scattering factor k Fire alarm boards: detection drawers detection lines Fire detectors: automatic press button Fire dampers Fire doors Dry sprinkler extinguishing systems (total failure) Dry sprinkler extinguishing system: (automatic actuation failure only) Wet sprinkler extinguishing systems (total failure) 1, Gas extinguishing systems (CO 2 ) Stationary fire pumps Figure 2 shows the contribution of the 10 main contributing fire zones or fire zone combinations. The main single contribution to the core damage frequency due to fire is /a (13 %) for a fire in the fire zone (ZF0258) containing the oil supply system of one main feedwater pump, the cables of the three pumps of the closed cooling water systems for nonnuclear plant and the cables of the start-up and shutdown pumps. In this case the only remaining heat sink is the emergency feedwater system. The second and third contribution are each /a (8 %) for fires ignited in the fire zone containing two loops inside containment and propagating to the adjacent fire zone containing the other two loops and vice versa (ZA0223/ZA0267 and ZA067/ZA0223). The obtained core damage frequency due to fire is /a. Compared with the core damage frequency of the initiating events of the Full Power PSA of /a this is a contribution of 12 %. Figure 2 shows the contribution of the 10 main contributing fire zones or fire zone combinations. The main single contribution to the core damage frequency due to fire is /a (13 %) for a fire in the fire zone (ZF0258) containing the oil supply system of one main feedwater pump, the cables of the three pumps of the closed cooling water systems for nonnuclear plant and the cables of the start-up and shutdown pumps. In this case the only remaining heat sink is the emergency feedwater system. The second and third contribution are each /a (8 %) for fires ignited in the fire zone containing two loops inside containment and propagating to the adjacent fire zone containing the other two loops and vice versa (ZA0223/ZA0267 and ZA067/ZA0223).

7 H.P.Berg page 7 of 8 Stage 1 Stage 2 Definition of fire zones Fire protection and component arrangement drawings Definition of fire zone combinations with connections Determination of fire frequencies Generic fire frequency Berry-method Screening of relevant fire zone combinations Determination of initiating events Plant knowledge, Component and cable routing lists Determination of initiating events Modeling of event trees + fault trees Full power PSA, System documents, Operating manual Modeling of event trees + fault trees Definition of boundery conditions Definition of boundery conditions Calculation of core damage frequencies Calculation of core damage frequencies Judgement of results Figure 1: Flow chart of the Fire PSA NPP Grafenrheinfeld 5 5 CONCLUDING REMARKS The first fire PSA that have been performed for NPP in Germany show that the contribution of plant internal fires to the total plant hazard state frequency is low and that these events do not represent - for the investigated plants - significant contributors to the plant hazard state frequency compared to the results from, e.g., US plants. However, the results of the fire PSA can be used to analyse in detail and assess findings of the deterministic part of the periodic safety review, to determine the necessity and urgency of safety improvements and to set priorities for fire protection improvement measures.

8 H.P.Berg page 8 of 8 Core damage frequency 1,00E-06 3,40E-07 1,00E-07 4,49E-08 2,60E-08 2,60E-08 2,14E-08 1,86E-08 1,72E-08 1,72E-08 1,00E-08 1,13E-08 1,13E-08 1,10E-08 Fire zone / fire zone combination Figure 2: Main contributing fire zones or fire zone combinations REFERENCES [1] Facharbeitskreis Probabilistische Sicherheitsanalyse für Kernkraftwerke. Methoden zur probabilistischen Sicherheitsanalyse für Kernkraftwerke, Dezember 1996, BfS-KT-16/97, Salzgitter [2] Facharbeitskreis Probabilistische Sicherheitsanalyse für Kernkraftwerke. Daten zur Quantifizierung von Ereignisablaufdiagrammen und Fehlerbäumen. April 1997, BfS-KT- 18/97, Salzgitter [3] Berg, H.P., Röwekamp, M., Experience from German Reliability Data for Fire Protection Measures in NPPs With Regard to Probabilistic Fire Safety Analyses. In: Proceedings 2nd International Conference Fire & Safety 97, Nuclear Engineering International / Wilmington Business Publishing, Dartford, pp [4] Berg, H.P., Röwekamp, M., Reliability Data Collection for Fire Protection Features, Kerntechnik, 65, No. 2-3, [5] Hristodulidis, A., Meyer, A.-W., Performance and Results of a Fire PSA for the Grafenrheinfeld 1300 MWe Nuclear Power Plant, Kerntechnik, 65, No. 2-3,